RESUMEN
The purpose of the research is to develop a statistical decision support algorithm for patients who may benefit from Adjuvant Cisplatin/Vinorelbine (ACT) and improve their survival rates. Genome-wide microarray data are used to identify feasible sets of genes and probe sets that constitute the gene signature. The data are available at the National Center for Biotechnology Information Gene Expression Omnibus (GSE14814). Preliminary studies have shown that high-risk patients who received ACT resulted in an improved prognosis. However, low-risk patients showed no benefit from ACT, and the treatment was possibly detrimental to the patient. Studies using tree-based ensemble statistical learning algorithms have shown that genomic markers could potentially identify a patient's risk factor and likelihood to benefit from ACT; however, it was noted that tree-based ensemble statistical learning algorithms do not provide an estimate of the strength of the treatment effect, nor is it possible to clearly identify subgroups of patients with similar responses to ACT treatment. Building on this idea, Accelerated Failure Time models are used to predict the probability of benefit from receiving chemotherapy or surgery only and provide a treatment recommendation for a new patient. We showed that regardless of whether the model recommended chemotherapy or surgery only, patients who followed the predicted treatment recommendation had significantly longer survival times than patients who did not. The proposed approach provides the likelihood of benefit for each treatment based on a small number of genomic biomarkers.